The Pentagon is officially moving its AI operations to OpenAI platforms after a public and messy divorce from Anthropic. The new agreement involves the deployment of large language models across classified networks, a move intended to streamline intelligence gathering, logistics, and strategic planning. This transition marks the end of the government’s experimentation with Anthropic’s Claude system, which is now banned from federal use.
The fallout was triggered by a disagreement over the limits of AI power. Anthropic insisted on keeping “kill switches” and ethical barriers in place that prevented the use of Claude in lethal autonomous weapons and domestic spying programs. The Pentagon, however, viewed these barriers as “strong-arming” by a private entity over the nation’s defense policy. This led to a stalemate that only ended when the President intervened.
President Trump’s involvement was a decisive factor in the shift. By publicly calling out Anthropic and directing all federal agencies to drop their services, he sent a clear message to the tech industry: cooperation with the Department of Defense is not optional. This “hardline” approach has forced AI companies to choose between their internal ethics boards and their ability to do business with the world’s largest spender.
OpenAI’s entry into this space is being watched closely by defense analysts. While Sam Altman claims the company will still adhere to strict ethical guidelines regarding autonomous weapons, the nature of “classified networks” makes public oversight difficult. OpenAI’s promise to the Pentagon includes a commitment to supporting national security while maintaining their core safety principles, a balancing act that will be tested in the coming months.
Anthropic’s departure from the federal space is being viewed by some as a blow to AI safety and by others as a necessary removal of an uncooperative vendor. Anthropic maintains that its restrictions never actually hindered any specific mission, suggesting that the dispute was more about the principle of government control over private technology. As OpenAI begins its integration, the eyes of the world are on how these “ethical” models will actually function in a theater of war.